GUC (Re: AW: AW: [HACKERS] Some notes on optimizer cost estimates)

On 2000-01-28, Tom Lane mentioned:

> Thomas Lockhart <lockhart(at)alumni(dot)caltech(dot)edu> writes:
> >> Added to TODO:
> >> * Unify configuration into one configuration file
>
> > ... and that is a good example of database design because?? ;)

This has more to do with software administration than database design or
databases in the first place. IMHO, o.c.

> One thing to consider while contemplating a grand unified config file
> (GUC?)

Darn, I was gonna suggest that name.

> is that much of this stuff needs to be settable per-client.
> It would be wrong to rip out whatever dynamic option-setting code
> there is. Cleaning it up and making a more uniform interface to the
> various options does sound like a good project though.

Nobody said anything about ripping out existing code. There just need to
be defaults settable somewhere without entering -o -F -f -B -q -R -n all
the time. I'm sure we can come up with something.

>
> I'd want to see a paper design for how things should work before any
> coding starts --- the existing methods do have some non-obvious
> advantages. For example, even something as grotty as the PGOPTIONS
> environment variable has its uses: you can pass options through to
> a backend without needing explicit cooperation from your client
> application.

So PGOPTIONS can be tied into the scheme.

--
Peter Eisentraut Sernanders väg 10:115
peter_e(at)gmx(dot)net 75262 Uppsala
http://yi.org/peter-e/ Sweden